[calls to trapsignal from assembler]

chuq@chuq.com said:
> is this patch still needed in the 3.x branch?

It is not a 100% complete fix, but it should save a crash in case
these code paths are ever hit. The siginfo generated is wrong,
fixing this would take much more work.

> it doesn't look it was ever checked in

Obviously not.. last checkin was 2003.
My problem was that I don't get any of these traps on
my boxes (which are R3000 DECstations, with FPUs, exclusively), so
I can't test it.

Hmm - looking at the code again it appears that one of the 3
patches can be exercised even on non-SOFTFLOAT systems by
synthesizing illegal FPU opcodes (which branch to the "ill" label
in fp.S, table driven)... some or most of these might be caught
by the hardware FPU earlier, don't know.

best regards
Matthias